Method Details
Details for method 'SERNet-Former (Berlin)'
Method overview
| name | SERNet-Former (Berlin) |
| challenge | pixel-level semantic labeling |
| details | Trained for 60 epochs |
| publication | SERNet-Former: Semantic Segmentation by Efficient Residual Network with Attention-Boosting Gates and Attention-Fusion Networks Serdar Erisen https://doi.org/10.48550/arXiv.2401.15741 |
| project page / code | |
| used Cityscapes data | fine annotations |
| used external data | ImageNet |
| runtime | n/a |
| subsampling | no |
| submission date | January, 2024 |
| previous submissions |
Average results
| Metric | Value |
|---|---|
| IoU Classes | 28.566 |
| iIoU Classes | 24.9568 |
| IoU Categories | 33.0628 |
| iIoU Categories | 35.0552 |
Class results
| Class | IoU | iIoU |
|---|---|---|
| road | 37.2875 | - |
| sidewalk | 19.6511 | - |
| building | 26.5876 | - |
| wall | 18.9359 | - |
| fence | 25.6888 | - |
| pole | 23.9124 | - |
| traffic light | 19.7378 | - |
| traffic sign | 20.7587 | - |
| vegetation | 40.3214 | - |
| terrain | 17.8677 | - |
| sky | 39.1853 | - |
| person | 25.0947 | 23.6003 |
| rider | 31.4719 | 21.1034 |
| car | 41.4672 | 46.153 |
| truck | 38.3355 | 22.2723 |
| bus | 36.9506 | 26.4666 |
| train | 31.0983 | 18.9342 |
| motorcycle | 21.1868 | 12.3927 |
| bicycle | 27.2152 | 28.7322 |
Category results
| Category | IoU | iIoU |
|---|---|---|
| flat | 35.6869 | - |
| nature | 38.9903 | - |
| object | 23.4698 | - |
| sky | 39.1853 | - |
| construction | 26.8217 | - |
| human | 26.6373 | 24.9839 |
| vehicle | 40.6484 | 45.1265 |